Elamur Research
☰
home / papers / 2025 / GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models

GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models

Written for the Research library (claude-opus-5-5) from the paper text. Original — arxiv.org/pdf/2508.06471

Summary

Open-source LLMs have trailed proprietary systems such as o3 and Claude Sonnet 4 when a single model must combine tool use, multi-step reasoning, and real-world software engineering. The GLM-4.5 Team (Zhipu AI & Tsinghua University) addresses this gap with GLM-4.5, a Mixture-of-Experts model with 355B total and 32B active parameters, and a lighter GLM-4.5-Air (106B total, 12B active). Both models can either reason step by step or answer directly. The design favors depth over width and adds extra attention heads, QK-Norm, and an MTP layer for speculative decoding. Training covers roughly 23T tokens. It begins with quality-bucketed general data and then shifts toward code and science. Mid-training follows, using repository-level code, synthetic reasoning traces, and long-context and agent data, while the context window grows from 4K to 128K. Post-training combines expert model iteration with reinforcement learning. Across 12 ARC benchmarks, GLM-4.5 ranks 3rd overall and 2nd on agentic tasks, despite having far fewer parameters than DeepSeek-R1 or Kimi K2. The open weights and evaluation toolkit give researchers a strong base for agent and coding work.

Key points

  • Headline scores for GLM-4.5: TAU-Bench 70.1%, BFCL v3 77.8%, BrowseComp 26.4%, AIME 24 91.0%, GPQA 79.1%, SWE-bench Verified 64.2%, and Terminal-Bench 37.5%.
  • Architecture: GLM-4.5 has 89 MoE layers, 160 experts with 8 active per token, and 96 attention heads with a 5120 hidden dimension. More heads did not lower training loss but did lift reasoning benchmarks such as MMLU and BBH.
  • Data recipe:
    • 15T-token general corpus, followed by a 7T code- and reasoning-focused corpus.
    • Mid-training with about 500B tokens of repo-level code, 500B of synthetic reasoning data, and 100B of long-context and agent data.
    • Web deduplication via SemDedup, with multilingual text drawn partly from Fineweb-2.
  • In a blind human evaluation on 100 hard translation cases (0–3 scale), GLM-4.5 scored 1.71, against 0.65 or less for Qwen-MT-plus, Qwen-MT-turbo, and Seed-X.
  • Weights are released at huggingface.co/zai-org/GLM-4.5, along with the glm-simple-evals toolkit for reproducing the benchmark results.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Natural language processing

Cited by

151 papers of our library cite it (of the 140673 whose bibliographies we read); 1 of them have a page here, 150 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2508.06471 · arXiv 2508.06471

@misc{zeng2025glm,
  title = {{GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models}},
  author = {Zeng, Aohan and Lv, Xin and Zheng, Qinkai and Hou, Zhenyu and Chen, Bin and Xie, Chengxing and Wang, Cunxiang and Yin, Da and Zeng, Hao and Zhang, Jiajie and Wang, Kedong and Zhong, Lucen and Liu, Mingdao and Lu, Rui and Cao, Shulin and Zhang, Xiaohan and Huang, Xuancheng and Wei, Yao and Cheng, Yean and An, Yifan and Niu, Yilin and Wen, Yuanhao and Bai, Yushi and Du, Zhengxiao and Wang, Zihan and Zhu, Zilin and Zhang, Bohan and Wen, Bosi and Wu, Bowen and Xu, Bowen and Huang, Can and Zhao, Casey and Cai, Changpeng and Yu, Chao and Li, Chen and Ge, Chendi and Huang, Chenghua and Zhang, Chenhui and Xu, Chenxi and Zhu, Chenzheng and Li, Chuang and Yin, Congfeng and Lin, Daoyan and Yang, Dayong and Jiang, Dazhi and Ai, Ding and Zhu, Erle and Wang, Fei and Pan, Gengzheng and Wang, Guo and Sun, Hailong and Li, Haitao and Li, Haiyang and Hu, Haiyi and Zhang, Hanyu and Peng, Hao and Tai, Hao and Zhang, Haoke and Wang, Haoran and Yang, Haoyu and Liu, He and Zhao, He and Liu, Hongwei and Yan, Hongxi and Liu, Huan and Chen, Huilong and Li, Ji and Zhao, Jiajing and Ren, Jiamin and Jiao, Jian and Zhao, Jiani and Yan, Jianyang and Wang, Jiaqi and Gui, Jiayi and Zhao, Jiayue and Liu, Jie and Li, Jijie and Li, Jing and Lu, Jing and Wang, Jingsen and Yuan, Jingwei and Li, Jingxuan and Du, Jingzhao and Du, Jinhua and Liu, Jinxin and Zhi, Junkai and Gao, Junli and Wang, Ke and Yang, Lekang and Xu, Liang and Fan, Lin and Wu, Lindong and Ding, Lintao and Wang, Lu and Zhang, Man and Li, Minghao and Xu, Minghuan and Zhao, Mingming and Zhai, Mingshu and Du, Pengfan and Dong, Qian and Lei, Shangde and Tu, Shangqing and Yang, Shangtong and Lu, Shaoyou and Li, Shijie and Li, Shuang and Shuang-Li and Yang, Shuxun and Yi, Sibo and Yu, Tianshu and Tian, Wei and Wang, Weihan and Yu, Wenbo and Tam, Weng Lam and Liang, Wenjie and Liu, Wentao and Wang, Xiao and Jia, Xiaohan and Gu, Xiaotao and Ling, Xiaoying and Wang, Xin and Fan, Xing and Pan, Xingru and Zhang, Xinyuan and Zhang, Xinze and Fu, Xiuqing and Zhang, Xunkai and Xu, Yabo and Wu, Yandong and Lu, Yida and Wang, Yidong and Zhou, Yilin and Pan, Yiming and Zhang, Ying and Wang, Yingli and Li, Yingru and Su, Yinpei and Geng, Yipeng and Zhu, Yitong and Yang, Yongkun and Li, Yuhang and Wu, Yuhao and Li, Yujiang and Liu, Yunan and Wang, Yunqing and Li, Yuntao and Zhang, Yuxuan and Liu, Zezhen and Yang, Zhen and Zhou, Zhengda and Qiao, Zhongpei and Feng, Zhuoer and Liu, Zhuorui and Zhang, Zichen and Wang, Zihan and Yao, Zijun and Wang, Zikang and Liu, Ziqiang and Chai, Ziwei and Li, Zixuan and Zhao, Zuodong and Chen, Wenguang and Zhai, Jidong and Xu, Bin and Huang, Minlie and Wang, Hongning and Li, Juanzi and Dong, Yuxiao and Tang, Jie},
  year = {2025},
  doi = {10.48550/arxiv.2508.06471},
  eprint = {2508.06471},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2508.06471}
}

References

52 references, 26 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. A. Abbas, K. Tirumala, D. Simig, S. Ganguli, and A. S. Morcos. Semdedup: Data-efficient learning at web-scale through semantic deduplication. arXiv preprint arXiv:2303.09540, 2023. find
  2. C. An, Z. Xie, X. Li, L. Li, J. Zhang, S. Gong, M. Zhong, J. Xu, X. Qiu, M. Wang, and L. Kong. Polaris: A post-training recipe for scaling reinforcement learning on advanced reasoning models, 2025. find
  3. Y. Bai, X. Lv, J. Zhang, H. Lyu, J. Tang, Z. Huang, Z. Du, X. Liu, A. Zeng, L. Hou, et al. Longbench: A bilingual, multitask benchmark for long context understanding. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 3119–3137,… find our card
  4. Y. Bai, S. Tu, J. Zhang, H. Peng, X. Wang, X. Lv, S. Cao, J. Xu, L. Hou, Y. Dong, J. Tang, and J. Li. LongBench v2: Towards deeper understanding and reasoning on realistic long-context multitasks. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: L… find our card
  5. M. Bavarian, H. Jun, N. Tezak, J. Schulman, C. McLeavey, J. Tworek, and M. Chen. Efficient training of language models to fill in the middle, 2022. find our card
  6. T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell, et al. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901, 2020. find
  7. A. Chen, A. Li, B. Gong, B. Jiang, B. Fei, B. Yang, B. Shan, C. Yu, C. Wang, C. Zhu, et al. Minimax-m1: Scaling test-time compute efficiently with lightning attention. arXiv preprint arXiv:2506.13585, 2025. find
  8. M. Chen, J. Tworek, H. Jun, Q. Yuan, H. P. D. O. Pinto, J. Kaplan, H. Edwards, Y. Burda, N. Joseph, G. Brockman, et al. Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374, 2021. find our card
  9. S. Cheng, Y. Bao, Q. Cao, L. Huang, L. Kang, Z. Liu, Y. Lu, W. Zhu, Z. Huang, T. Li, et al. Seed-x: Building strong multilingual translation llm with 7b parameters. arXiv preprint arXiv:2507.13618, 2025. — [arXiv:2507.13618](https://arxiv.org/abs/2507.13618) find
  10. K. Deshpande, V. Sirdeshmukh, J. B. Mols, L. Jin, E.-Y. Hernandez-Cardona, D. Lee, J. Kritz, W. E. Primack, S. Yue, and C. Xing. Multichallenge: A realistic multi-turn conversation evaluation benchmark challenging to frontier llms. In Findings of the Association for Computational Linguistics: ACL 20… find
  11. H. Ding, Z. Wang, G. Paolini, V. Kumar, A. Deoras, D. Roth, and S. Soatto. Fewer truncations improve language modeling. In Proceedings of the 41st International Conference on Machine Learning, pages 11030–11048, 2024. find
  12. F. Gloeckle, B. Y. Idrissi, B. Rozière, D. Lopez-Paz, and G. Synnaeve. Better & faster large language models via multi-token prediction. arXiv preprint arXiv:2404.19737, 2024. find
  13. D. Guo, D. Yang, H. Zhang, J. Song, R. Zhang, R. Xu, Q. Zhu, S. Ma, P. Wang, X. Bi, et al. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948, 2025. find our card
  14. D. Hendrycks, C. Burns, S. Kadavath, A. Arora, S. Basart, E. Tang, D. Song, and J. Steinhardt. Measuring mathematical problem solving with the math dataset. In Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2). find our card
  15. A. Henry, P. R. Dachapally, S. Pawar, and Y. Chen. Query-key normalization for transformers, 2020. find
  16. C.-P. Hsieh, S. Sun, S. Kriman, S. Acharya, D. Rekesh, F. Jia, and B. Ginsburg. Ruler: What’s the real context size of your long-context language models? In First Conference on Language Modeling. find our card
  17. S. Hu, Y. Tu, X. Han, G. Cui, C. He, W. Zhao, X. Long, Z. Zheng, Y. Fang, Y. Huang, et al. Minicpm: Unveiling the potential of small language models with scalable training strategies. In First Conference on Language Modeling. find our card
  18. A. Jaech, A. Kalai, A. Lerer, A. Richardson, A. El-Kishky, A. Low, A. Helyar, A. Madry, A. Beutel, A. Carney, et al. Openai o1 system card. arXiv preprint arXiv:2412.16720, 2024. find our card
  19. N. Jain, K. Han, A. Gu, W.-D. Li, F. Yan, T. Zhang, S. Wang, A. Solar-Lezama, K. Sen, and I. Stoica. Livecodebench: Holistic and contamination free evaluation of large language models for code. In The Thirteenth International Conference on Learning Representations. find our card
  20. C. E. Jimenez, J. Yang, A. Wettig, S. Yao, K. Pei, O. Press, and K. Narasimhan. Swe-bench: Can language models resolve real-world github issues? arXiv preprint arXiv:2310.06770, 2023. find our card
  21. K. Jordan, Y. Jin, V. Boza, Y. Jiacheng, F. Cecista, L. Newhouse, and J. Bernstein. Muon: An optimizer for hidden layers in neural networks, 2024. URL https://kellerjordan.github.io/posts/muon, 6. ↗
  22. A. Joulin, E. Grave, P. Bojanowski, and T. Mikolov. Bag of tricks for efficient text classification. In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers, pages 427–431. Association for Computational Linguistics, April… find
  23. A. Liu, B. Feng, B. Xue, B. Wang, B. Wu, C. Lu, C. Zhao, C. Deng, C. Zhang, C. Ruan, et al. Deepseek-v3 technical report. arXiv preprint arXiv:2412.19437, 2024. find
  24. J. Liu, J. Su, X. Yao, Z. Jiang, G. Lai, Y. Du, Y. Qin, W. Xu, E. Lu, J. Yan, et al. Muon is scalable for llm training. arXiv preprint arXiv:2502.16982, 2025. find
  25. M. Luo, S. Tan, J. Wong, X. Shi, W. Y. Tang, M. Roongta, C. Cai, J. Luo, L. E. Li, R. A. Popa, and I. Stoica. Deepscaler: Surpassing o1-preview with a 1.5b model by scaling rl. https://pretty-radio-b75.notion.site/DeepScaleR-Surpassing-O1-Preview-with-a-1-5B-Model-by-Scaling-RL-19681902c1468005bed8c… ↗
  26. S. G. Patil, H. Mao, C. Cheng-Jie Ji, F. Yan, V. Suresh, I. Stoica, and J. E. Gonzalez. The berkeley function calling leaderboard (bfcl): From tool use to agentic evaluation of large language models. In Forty-second International Conference on Machine Learning, 2025. find
  27. G. Penedo, H. Kydlíček, V. Sabolčec, B. Messmer, N. Foroutan, A. H. Kargaran, C. Raffel, M. Jaggi, L. Von Werra, and T. Wolf. Fineweb2: One pipeline to scale them all–adapting pre-training data processing to every language. arXiv preprint arXiv:2506.20920, 2025. find
  28. L. Phan, A. Gatti, Z. Han, N. Li, J. Hu, H. Zhang, C. B. C. Zhang, M. Shaaban, J. Ling, S. Shi, et al. Humanity’s last exam. arXiv preprint arXiv:2501.14249, 2025. find our card
  29. Y. Qin, T. Zhang, Y. Shen, W. Luo, Y. Zhang, Y. Qiao, Z. Zhou, W. Zhang, B. CUI, et al. Sysbench: Can llms follow system message? In The Thirteenth International Conference on Learning Representations, 2024. find
  30. D. Rein, B. L. Hou, A. C. Stickland, J. Petty, R. Y. Pang, J. Dirani, J. Michael, and S. R. Bowman. Gpqa: A graduate-level google-proof q&a benchmark. In First Conference on Language Modeling, 2024. find our card
  31. Z. Shao, P. Wang, Q. Zhu, R. Xu, J. Song, X. Bi, H. Zhang, M. Zhang, Y. Li, Y. Wu, et al. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300, 2024. find our card
  32. D. Su, K. Kong, Y. Lin, J. Jennings, B. Norick, M. Kliegl, M. Patwary, M. Shoeybi, and B. Catanzaro. Nemotron-cc: Transforming common crawl into a refined long-horizon pretraining dataset. arXiv preprint arXiv:2412.02595, 2024. find
  33. G. Team, R. Anil, S. Borgeaud, J.-B. Alayrac, J. Yu, R. Soricut, J. Schalkwyk, A. M. Dai, A. Hauth, K. Millican, et al. Gemini: a family of highly capable multimodal models. arXiv preprint arXiv:2312.11805, 2023. find our card
  34. K. Team, Y. Bai, Y. Bao, G. Chen, J. Chen, N. Chen, R. Chen, Y. Chen, Y. Chen, Y. Chen, et al. Kimi k2: Open agentic intelligence. arXiv preprint arXiv:2507.20534, 2025. find our card
  35. T. T.-B. Team. Terminal-bench: A benchmark for ai agents in terminal environments, Apr 2025. find
  36. M. Tian, L. Gao, S. Zhang, X. Chen, C. Fan, X. Guo, R. Haas, P. Ji, K. Krongchon, Y. Li, et al. Scicode: A research coding benchmark curated by scientists. Advances in Neural Information Processing Systems, 37:30624–30650, 2024. find
  37. H. Touvron, T. Lavril, G. Izacard, X. Martinet, M.-A. Lachaux, T. Lacroix, B. Rozière, N. Goyal, E. Hambro, F. Azhar, et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023. find
  38. K. Vodrahalli, S. Ontanon, N. Tripuraneni, K. Xu, S. Jain, R. Shivanna, J. Hui, N. Dikkala, M. Kazemi, B. Fatemi, R. Anil, E. Dyer, S. Shakeri, R. Vij, H. Mehta, V. Ramasesh, Q. Le, E. Chi, Y. Lu, O. Firat, A. Lazaridou, J.-B. Lespiau, N. Attaluri, and K. Olszewska. Michelangelo: Long context evalua… find
  39. F. Wan, W. Shen, S. Liao, Y. Shi, C. Li, Z. Yang, J. Zhang, F. Huang, J. Zhou, and M. Yan. Qwenlong-l1: Towards long-context large reasoning models with reinforcement learning. arXiv preprint arXiv:2505.17667, 2025. find
  40. L. Wang, H. Gao, C. Zhao, X. Sun, and D. Dai. Auxiliary-loss-free load balancing strategy for mixture-of-experts. arXiv preprint arXiv:2408.15664, 2024. — [arXiv:2408.15664](https://arxiv.org/abs/2408.15664) find
  41. S. Wang, L. Yu, C. Gao, C. Zheng, S. Liu, R. Lu, K. Dang, X. Chen, J. Yang, Z. Zhang, et al. Beyond the 80/20 rule: High-entropy minority tokens drive effective reinforcement learning for llm reasoning. arXiv preprint arXiv:2506.01939, 2025. find our card
  42. X. Wang, B. Li, Y. Song, F. F. Xu, X. Tang, M. Zhuge, J. Pan, Y. Song, B. Li, J. Singh, H. H. Tran, F. Li, R. Ma, M. Zheng, B. Qian, Y. Shao, N. Muennighoff, Y. Zhang, B. Hui, J. Lin, R. Brennan, H. Peng, H. Ji, and G. Neubig. Openhands: An open platform for AI software developers as generalist agen… find our card
  43. Y. Wang, X. Ma, G. Zhang, Y. Ni, A. Chandra, S. Guo, W. Ren, A. Arulraj, X. He, Z. Jiang, et al. Mmlu-pro: A more robust and challenging multi-task language understanding benchmark. Advances in Neural Information Processing Systems, 37:95266–95290, 2024. find our card
  44. J. Wei, N. Karina, H. W. Chung, Y. J. Jiao, S. Papay, A. Glaese, J. Schulman, and W. Fedus. Measuring short-form factuality in large language models, 2024. find our card
  45. J. Wei, Z. Sun, S. Papay, S. McKinney, J. Han, I. Fulford, H. W. Chung, A. T. Passos, W. Fedus, and A. Glaese. Browsecomp: A simple yet challenging benchmark for browsing agents. arXiv preprint arXiv:2504.12516, 2025. find
  46. Z. Xi, Y. Ding, W. Chen, B. Hong, H. Guo, J. Wang, D. Yang, C. Liao, X. Guo, W. He, et al. Agentgym: Evolving large language model-based agents across diverse environments. arXiv preprint arXiv:2406.04151, 2024. find
  47. A. Yang, A. Li, B. Yang, B. Zhang, B. Hui, B. Zheng, B. Yu, C. Gao, C. Huang, C. Lv, et al. Qwen3 technical report. arXiv preprint arXiv:2505.09388, 2025. find our card
  48. S. Yao, N. Shinn, P. Razavi, and K. Narasimhan. tau-bench: A benchmark for tool-agent-user interaction in real-world domains. arXiv preprint arXiv:2406.12045, 2024. find our card
  49. Q. Yu, Z. Zhang, R. Zhu, Y. Yuan, X. Zuo, Y. Yue, W. Dai, T. Fan, G. Liu, L. Liu, et al. Dapo: An open-source llm reinforcement learning system at scale. arXiv preprint arXiv:2503.14476, 2025. find our card
  50. A. Zeng, X. Liu, Z. Du, Z. Wang, H. Lai, M. Ding, Z. Yang, Y. Xu, W. Zheng, X. Xia, et al. Glm-130b: An open bilingual pre-trained model. In The Eleventh International Conference on Learning Representations. find our card
  51. Z. Zhang, L. Lei, L. Wu, R. Sun, Y. Huang, C. Long, X. Liu, X. Lei, J. Tang, and M. Huang. Safetybench: Evaluating the safety of large language models with multiple choice questions. arXiv preprint arXiv:2309.07045, 2023. find our card
  52. J. Zhou, T. Lu, S. Mishra, S. Brahma, S. Basu, Y. Luan, D. Zhou, and L. Hou. Instruction-following evaluation for large language models. arXiv preprint arXiv:2311.07911, 2023. find our card